Re: [Anuga-user] Problem with runtests.py / _intel_fast_memset issue

[Christos A. <alt...@gm...>]

Hi Steve,

Please find attached the build.log according to the instructions you
provided.

I found a solution to my problem. I installed a local conda environment and
I managed to setup everything properly. I can run ANUGA both serially and
parallel. I am not sure about efficiency yet but I will work on this as
well.

The build.log might not be the best since I was not very careful. In order
to get it, I isolated the conda environment and I tried to setup anuga with
my standard bashrc profile. I think the solution would be to setup anuga
with intel compilers as you suggested. I will discuss about it with the
cluster guys and I will try it. I guess it will take some trials but I will
figure it out. I will let you know of any other issues that might appear.

Thank you very much for your help.

Best
Christos

On Fri, Aug 2, 2019 at 7:20 PM Christos Altantzis <alt...@gm...>
wrote:

> Hi Steve,
>
> Thanks a lot for the immediate response. Thank you for the links as well
> as the suggestions. I was already aware of the two first links and while I
> forced loading only gcc, I still  have the same issues.
>
> Currently I am trying to setup everything through a local conda
> environment. After I complete this try, I will isolate the conda
> environment and load my previous python setup in order to try what you
> suggested.
>
> I will let you know of my progress.
>
> Notifying the admins on Friday afternoon, August 2 in Europe!! My chances
> are not that high. I will let them know on Monday.
>
> Thanks a lot again.
>
> Best
> Christos
>
> On Fri, Aug 2, 2019 at 3:49 PM Stephen Roberts <ste...@an...>
> wrote:
>
>> Hi Christos,
>>
>>
>> It would be interesting to see the output of
>>
>>
>> python setup.py clean
>>
>> python setup.py build 2>&1 | tee build.log
>>
>>
>> The content of build.log might give us a hint as to where the intel
>> command is coming from.
>>
>>
>> I suppose you have talked to your system admin people?
>>
>>
>> Some interesting discussions
>>
>>
>> https://mail.python.org/pipermail//scikit-image/2015-October/004333.html
>>
>>
>> https://groups.google.com/forum/#!topic/scikit-image/iSgNGBG0ZqU
>>
>>
>> and
>>
>>
>>
>> https://christoph-conrads.name/building-numpy-and-scipy-with-intel-compilers-and-intel-mkl/
>>
>>
>> It has been a while but we have compiled anuga using the intel compilers,
>> so this third link might be an option.
>>
>>
>> Cheers
>>
>> Steve
>>
>>
>> ==============================
>> Stephen Roberts
>> Undergraduate Convenor
>> Mathematical Sciences Institute
>> Room 4.74 Hanna Neumann Building #145
>> The Australian National University
>> Canberra, ACT 2600 AUSTRALIA
>> Ph: +61 2 61254445
>> CRICOS: 00120C
>>
>> ------------------------------
>> *From:* Christos Altantzis <alt...@gm...>
>> *Sent:* Friday, 2 August 2019 8:48:48 PM
>> *To:* Anu...@li... <Anu...@li...>
>> *Subject:* [Anuga-user] Problem with runtests.py / _intel_fast_memset
>> issue
>>
>> Hi Group,
>>
>> I have been using anuga for more than a year now and I have to admit that
>> the developers have done an excellent job. I really thank you for this.
>> Until now I have been working on local machines and I never had a problem.
>>
>> Currently, I am trying ti install ANUGA on a cluster where I do not have
>> superuser rights, so I am trying to install the code from source. I have
>> managed to install all the dependencies and all of them seem to be working
>> properly. Build_all.py works without problems and the installation seems to
>> be fine. Then when I am trying to import anuga or run the runtests.py I get
>> the following error:
>>
>> Building, see build.log...
>> Build OK
>> Traceback (most recent call last):
>>   File "runtests.py", line 402, in <module>
>>     main(argv=sys.argv[1:])
>>   File "runtests.py", line 201, in main
>>     __import__(PROJECT_MODULE)
>>   File
>> "/cluster/home/caltantz/Simulations/ANUGA/anuga_core/build/testenv/lib64/python2.7/site-packages/anuga/__init__.py",
>> line 69, in <module>
>>     from anuga.shallow_water.shallow_water_domain import Domain
>>   File
>> "/cluster/home/caltantz/Simulations/ANUGA/anuga_core/build/testenv/lib64/python2.7/site-packages/anuga/shallow_water/shallow_water_domain.py",
>> line 100, in <module>
>>     from anuga.abstract_2d_finite_volumes.generic_domain \
>>   File
>> "/cluster/home/caltantz/Simulations/ANUGA/anuga_core/build/testenv/lib64/python2.7/site-packages/anuga/abstract_2d_finite_volumes/generic_domain.py",
>> line 22, in <module>
>>     from quantity import Quantity
>>   File
>> "/cluster/home/caltantz/Simulations/ANUGA/anuga_core/build/testenv/lib64/python2.7/site-packages/anuga/abstract_2d_finite_volumes/quantity.py",
>> line 30, in <module>
>>     from anuga.fit_interpolate.fit import fit_to_mesh
>>   File
>> "/cluster/home/caltantz/Simulations/ANUGA/anuga_core/build/testenv/lib64/python2.7/site-packages/anuga/fit_interpolate/fit.py",
>> line 33, in <module>
>>     from anuga.fit_interpolate.general_fit_interpolate import
>> FitInterpolate
>>   File
>> "/cluster/home/caltantz/Simulations/ANUGA/anuga_core/build/testenv/lib64/python2.7/site-packages/anuga/fit_interpolate/general_fit_interpolate.py",
>> line 26, in <module>
>>     from anuga.utilities.sparse import Sparse, Sparse_CSR
>>   File
>> "/cluster/home/caltantz/Simulations/ANUGA/anuga_core/build/testenv/lib64/python2.7/site-packages/anuga/utilities/sparse.py",
>> line 335, in <module>
>>     from sparse_ext import csr_mv
>> ImportError:
>> /cluster/home/caltantz/Simulations/ANUGA/anuga_core/build/testenv/lib64/python2.7/site-packages/anuga/utilities/sparse_ext.so:
>> undefined symbol: _intel_fast_memset
>>
>> The problem is this _intel_fast_memset. I made sure that I do not load
>> any icc in my bashrc and that only gcc is used, and still I see this
>> problem. Is it possible that this happens due to the fact that some of the
>> dependencies might have been built with icc?? Have you encountered such an
>> issue in the past?? I am struggling to resolve this issue but I am out of
>> ideas.
>>
>> If anything else fails I will try to install locally conda and do
>> everything through conda, but if somebody has some experience with a
>> similar problem I would appreciate sharing his/her experience.
>>
>> Best
>> Christos
>>
>